Lidar Point Cloud Segmentation for Ground and Vegetation Separation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional algorithms for lidar point cloud segmentation in autonomous vehicles struggle to accurately identify and differentiate between objects such as ground and vegetation, leading to errors, especially in complex environments like groups of pedestrians or non-convex objects.
Innovation Solution
A neural network-based lidar data segmentation system is employed, which receives input features from lidar data and outputs probabilities for each point to classify it as ground, vegetation, or other objects, allowing for accurate segmentation by assigning labels and reducing errors through exclusion of misclassified points.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional algorithms based on human-generated rules are used for lidar point cloud segmentation, then the system can identify objects and their locations, but the segmentation is not robust to variations in the driving environment and results in errors with object identification
Solution Approach 1:
The patent replaces conventional rule-based algorithms with a neural network-based machine learning system. The neural network learns segmentation patterns from training data and automatically adapts to different environmental conditions without requiring manual rule adjustments, thereby improving both reliability and adaptability of object identification.
Solution Approach 2:
The system changes the fundamental parameter of segmentation from fixed human-generated rules to dynamic parameters learned from data. The neural network adjusts its internal parameters (weights and biases) based on training data, enabling it to handle variations in driving environments such as different lighting conditions, weather, and object configurations.
2Adaptability or versatility
If conventional segmentation algorithms are used, then the system can process lidar data, but it performs poorly when certain types of objects are present such as groups of closely-spaced pedestrians, vegetation, or non-convex objects
Solution Approach 1:
The neural network replaces conventional segmentation algorithms that struggle with complex objects. It learns to identify object boundaries and characteristics by processing training examples of challenging scenarios such as closely-spaced pedestrians, vegetation, and non-convex objects, achieving better measurement precision in these difficult cases.
Solution Approach 2:
The system performs preliminary training on diverse and complex object scenarios before deployment. The neural network is trained in advance on datasets containing groups of closely-spaced pedestrians, vegetation, and non-convex objects, enabling it to handle these complex situations accurately when encountered in real driving environments.
Data Source
AI summary
An autonomous vehicle is described herein. The autonomous vehicle includes a lidar sensor system. The autonomous vehicle additionally includes a computing system that executes a lidar segmentation system, wherein the lidar segmentation system is configured to identify objects that are in proximity to the autonomous vehicle based upon output of the lidar sensor system. The computing system further includes a deep neural network (DNN), where the lidar segmentation system identifies the objects in proximity to the autonomous vehicle based upon output of the DNN.